S1E10
The Case for Cash Flow Underwriting
November 20, 2024
Cash flow underwriting: it's the next big thing. And so it has been for nearly a decade, admits Nikki Cross. Nikki is the Senior Director of Data Science Solutions at Nova Credit, an alternative credit data platform. She joins Shawn to make the case for cash flow underwriting. Why do lenders struggle with incorporating cash flow data, and how can they speed up the process? Why should lenders make the effort? By the end of the episode, you'll be excited about cash flow underwriting...the next big thing.
Find Nikki and Ensemblex on LinkedIn.
Hosted by: Shawn Budde
Guest: Nikki Cross
Produced by: Meagan LeBlanc
Theme Music by: Brad Frank
Transcript
Shawn
Hello, this is Shawn Budde of Ensemblex and this is The Ensemblex Exchange Podcast. Today I'm talking with Nikki Cross of Nova Credit about cashflow underwriting. Nikki has spent her career in data science and financial services with a focus on the alternative data space. She has spent 13 years on the lender side of the business, including nearly a decade at Capital One. She's also spent over seven years on the credit bureau side of the fence, including being part of L2C's acquisition into TransUnion. In 2021, she joined Nova Credit to build out their solutions consulting function and drive their adoption as a consumer permissioned credit bureau. Welcome, Nikki.
Nikki
Thank you so much for having me, Shawn.
Shawn
My pleasure. So Nikki, you are our second guest to have visited all 50 states. So let's start there. I don't think that happens accidentally. So when did you start that mission? When did you finish that mission?
Nikki
I didn't really make it a goal to visit all 50 states until probably 2016, 2017, because at that point I was at TransUnion. I was traveling almost every week for work. And so you suddenly just start hitting, there was a year where I hit like 15 new states. And so it's like, well, if I'm going to be on the road all the time. And then I started to like do those things. I will fully admit probably six or seven are kind of those cheat states where I was like, New Hampshire and I drove across the border had dinner in Vermont and came back those sorts of things so I will not say that I have seen all and experienced the totality of all 50 but I have at least had a meal and and they're probably like I said six or seven of those.
Shawn
We had this debate. I was in Europe with my brother and my dad, and we were debating the definition of having visited a country. And we settled on dinner or I should say a meal or a night. For 50 states, so I think you gotta spend the night, Nikki. I'm not gonna give you full credit.
Nikki
I might have then probably 10 that I haven't actually spent the night in. So that's now, that's a new goal that I should establish and I'll have to go recalculate that list for myself.
Shawn
I'm sitting at 49, so Alaska is yet to go.
Nikki
I did make it to Alaska. My last state was actually Rhode Island of all things. It had to be kind of a concerted effort because you just don't happen to find yourself in Rhode Island very often.
Shawn
That is absolutely true. Yeah, I spent a night in Providence. That got me off the hook there. So you recently, I think, went rafting through the Grand Canyon.
Nikki
I did, yeah.
Shawn
And how long was that?
Nikki
So I was on the river for two weeks, you can be on the river for different periods of time depending upon how far you're going. You can do kind of a half canyon hike out or hike in and then do bottom canyon. You also depending on the type of boat you're on. So if you're on a people powered boat then it's a slower trip than if you're on a motorboat for example. So I was on people powered boats and did the full canyon. So 14 days on the river.
Shawn
Yeah, my mom did it twice and loved it so much that she wanted to take my daughters. And so she told them when they turned 12, she would take each of them on the raft trip and none of them opted into the raft trip, they each got a vacation out of her rafting down the Grand Canyon. All right. So why don't we dive in? So your, your background is statistics. I think interestingly you went to Georgia Tech, graduated in 20 years, decided you hadn't had enough and you went back. So I'm curious about that. Why go back 20 years into your career?
Nikki
That's right.
Nikki
Yeah, it was one of those things where if you think about kind of the time I graduated from my undergrad in 01, I graduated from my master's almost to the day 20 years later in 2021. It's been a huge shift in data science since then, not the least of which is that we call it data science and it used to be statistics. And so everybody got a title change. But really like that transition of the software people were working and the techniques that they were using, all of those things had kind of shifted while I was working on the credit bureau side of things and was acting more as a consultant and not literally building some of those models and kind of hands-on keyboard in those ways. And I felt like that was a real gap in my understanding because I wanted to be able to do code reviews for my team. I wanted to be able to talk appropriately about some of the new techniques that have been evolving in the way that we're using those. And so for me, it was a real forcing mechanism to be able to get some of those experiences. And I always said like, it was the job I would need for me to hire me for an entry level position, right? Like it was that kind of education that was really foundational that I could have gotten in a number of other ways. There are a lot of really great ways to do that. But for me going back and getting, going through a more organized program was the right construct that I needed.
Shawn
Yeah, that's interesting. So you mentioned you were on the credit bureau side at the time you went back. So talk to me about how is it different to be on that side of the equation versus on the consuming, I guess, side of the credit bureau equation.
Nikki
Yeah, it's almost like speaking two different languages because there are a lot of reasons that credit bureaus operate the way they do, not the least of which is that they have decades of history and decades of technical infrastructure, decades of things that have been set up in those ways. And so when you know why things work the way they do, you can engage with that data and those experiences very, very differently. And so I think it's a really interesting engagement model, right? So when you're working on the lender side, you probably have a small group of folks that you're working with relatively closely. As a consultant on the credit bureau side, you're working with a ton of people, a bunch of different organizations, but on a more shallow level. So how deep you get, kind of what you get to do, your experiences in implementing, all of those things look very, very different on each of those sides of the fence. But it does mean that you can connect the dots when you've been on both sides of those worlds.
Shawn
So Nova Credit does not have decades of infrastructure. So how has Nova Credit approached this differently than the way the credit bureaus that have been around for decades?
Nikki
Yeah, absolutely. So the interesting thing about Nova Credit is that we're not really actually sitting on a database at all. So when you think about the traditional credit bureaus, they've got decades of history for hundreds of millions of people. And what Nova Credit is, is focused on consumer permissioned data. So we are helping, for example, on our international data side, we're helping an immigrant who has potentially recently come to the US, pull a copy of their report from the UK, India, Nigeria, and bring that to the US.
We don’t own those reports that that consumer is pulling, we're helping them get that from that home country report. And similarly in the cashflow underwriting space that we'll talk about today, we're not sitting on a database of all of the consumer's depository information. We're helping that consumer permission it. We add intelligence scores, attributes on top of that data, and then we're using that in real time. So you have to think about the data very, very differently. How we organize, how we access the data is all very different because it must happen in real time and you don't have that data store that's sitting that legacy database and infrastructure there. So it changes how you access the data, it changes the analytics you can do, it changes the way you build a credit policy. All of those things are really unusual for an organization like Nova.
Shawn
Let's actually take a slight detour into that international credit bureau aspect. How does that work? I mean, are you creating a uniform set of variables because you're pulling from all these different places or is every bureau looking different and you're just kind of a clearinghouse?
Nikki
It's a little bit of a combination of the two. So we're working with existing credit bureaus in the same way that US consumers were accustomed to the ability to pull their report once a year, for example. We help that consumer go to UK, India, Nigeria. We support about 20 bureaus currently live and are always adding additional bureaus. And each of those reports is going to look very, very different, right? The language that it's in, the currency that it's in, the amount of data that's stored. There is, for example, a concept of negative only bureaus.
If you've paid your credit card as expected, that will never ever get reported to those bureaus. Only if you have a delinquency or some sort of a default, do you even appear at that bureau, for example. And so what Nova Credit does is we handle the logistics of literally getting a copy of that report. We digest it into a uniform report, regardless where it comes from, with a suite of attributes and a score. And then we give a copy of that report to the consumer as well as to the lender.
And so that enables some of our lending partners to be able to make decisions based on a leveled score, regardless of where the report came from, a uniform set of attributes, regardless of where that report came from. And now suddenly that opens up consumers who may be new to the US, have been here, maybe just have recently moved, have been here even six or 12 months. It takes time to establish that footprint, but you've got five or 10 years of history, right? You know that this person successfully paid a mortgage for 12 years.
Why are we ignoring that information? And now suddenly you've got a really rich source of population, right? Because most of our population growth in the US is gonna come from immigrants over the next 30 years. And also you're not just treating these folks because they have moved as if they didn't have a life before. And that's just a really, it's difficult enough for the immigrant experience and all of the things that you have to do. And so the ability to more easily get a credit card, to get a cell phone, to be able to establish your life when you're coming here for school, for higher education, for work, you know, pretending that they don't have decades of history is a silly thing to do. And so leveling that playing field a bit.
Shawn
Got it. Okay, so, Nikki, cashflow underwriting is, I wanna say it's a new concept, but maybe it's actually a really old concept. What is it? Just describe cashflow underwriting for me.
Nikki
Sure, to your point that it is a very old concept that is new again, I would say. If you imagine like back before there were credit bureaus and in the 40s, 50s, 60s, if you'd walked into a branch of your local bank and said, hey, I need to borrow money to buy a house. The first thing they would do is look at how much deposit you had and what your status had been with that bank. And that would probably be the experience you would have had sitting in that branch with a banker. And so that idea of understanding what someone's depository information looks like is complementary to what we see in a traditional credit bureau. And what's happened is that with the advent of a lot of the aggregators who are now there, the ability to access that data in a different way, right? It's, if you think about some of the biggest banks, they've always been able to use it in lending because they typically had those balances and they understood what a consumer was looking at from the depository side. Someone who was only a lender wouldn't have access to that and were really at a strategic disadvantage over time.
Now, with things like 1033 and the ability to consumer permission this data, you can level that playing field. And now suddenly you have the intelligence of understanding when a consumer is living paycheck to paycheck, for example, and what sort of assets they may have that go beyond what you're seeing in the credit report.
Shawn
Yeah, so that's an interesting, I guess, return, so to speak. I suppose at one point in time, people had a bank, right? And all their data to your point kind of sat there. And then over time, it maybe got dispersed as they took debt in one place and deposits in another. Before we move on to that, though, I want to understand what is a 1033?
Nikki
1033 is part of the regulation that CFPB is working on. Right now it is, shall we say, sometimes challenging to work with some of the banks who limit the amount of data that they let outside their doors. That understanding of is the data owned by the consumer? Is it owned by the bank, for example? There's a lot of nuance there in terms of what a consumer can permission, how much data goes out the door and things like that. And so 1033 is the regulation that's being worked on, reviewed currently that will start to normalize and in some cases probably mandate the data that a bank must allow to go out. And so it's very similar to in Europe, there are a lot of open banking regulations that have been put in place to be able to democratize access to that data where the consumers have given that approval.
Shawn
Right, so we've worked with it in the UK, we've worked with it in Brazil and also in India. I have to say, really, it is a game changer and it really enables everyone else to kind of compete on that level playing field. So going back to the cashflow underwriting, I think one of the variables that is intuitively obvious in underwriting and in my experience is empirically poor is debt to income.
And cashflow underwriting at some level sounds to me like a version of debt to income. So can you explain why is it different? I guess do you have the same bias against debt to income that I do? And why does cashflow underwriting work when DTI maybe doesn't?
Nikki
Yeah, I think anything you try to boil down to a single number, right? Debt to income, things like that. I think that becomes a little bit problematic. I think one thing that debt to income also looks at is generally when you're calculating that, you're basing that upon the payments that are required for credit products only, right? And then you're comparing that to an income. You're not seeing other expenses. Somebody's rent, for example, is generally not being included there. It's not in a verified way, things like that, you're missing out on a lot of that picture. And so what you start to see when you have the ability to take cash flow, that depository data, along with the credit report, is you get a much more holistic view of a consumer's finances. And so that cuts both ways, right? There are places where if someone has recently been laid off, you suddenly understand that that cash flow and that income has gone to zero. And that, you're going to be able to identify in a depository account where their paycheck was being deposited before you're going to see the time it has taken them to go delinquent on a credit card or an auto payment and make it to the credit report, for example. Similarly, if I've gotten a new promotion, if I'm making more money, I've got a new job, you're going to see that increase in income start to be reflected. And we live in a world where half of American households live paycheck to paycheck, including a third of high income households. And I, as personal story, grew up in rural Georgia and I live in Santa Monica, California.
A dollar of income looks very different in those two locations. And so income in and of itself just isn't enough, right? We have to understand what are my spending habits? What sort of safety net do I have? Am I one medical bill, one flat tire away from not being able to pay my bills? Or have I kind of built up that cushion and created a world where I do have some flexibility? And all of those things really become a much more interesting and a richer picture of who that consumer is when we can look at this in a total package.
Shawn
Okay, so you say, you know, one flaw of DTIs, it's a single metric, trying to do a lot. Cashflow underwriting sounds to me like a single metric, like how much money am I left over with at the end of every month? But I assume you're saying it's more than that. So how do I extract many variables out of that one set of information?
Nikki
Yeah, so if you think about like what you would see if you look into your own checking account, there are a couple of different ways that Nova Credit aggregates that information. Our package within Cash Atlas has about 1200 attributes that we make available to lenders as well as a score. So it is a bounty of information that we can derive there. And we categorize those, each of those line items that you see in that raw data into three different categories, income, expenses, and then assets, which is kind of the net of income and expenses.
Now, we do a bunch of combinations in terms of min and max as trends over time, how long we look at that data, three, six, nine, 12 months, for example. So when I say there are 1200 attributes, there are some basic concepts that you'll see with a bunch of combinations there. But what that really means is, again, income in and of the self isn't enough. So we look at things like the number of days that you've had a balance under $100, $500, under $1 ,000. We start to understand what the stability of that consumer is in a way that accounts for both the depository as well as the expenses there. We look at expenses through a lens of recurring and semi-recurring, for example. So when we start to understand there, how much is that consumer kind of operating in a way that they're accustomed to versus what are the shocks in the system? And so as we start to look at things in a different way and start to kind of consider those line items differently than you probably would from a personal financial management, we're not looking at groceries versus fast food.
But looking at those from a risk lens to understand which are the attributes that are most indicative of that consumer's risk, then that's when we get that really rich set of attributes and a score to be able to actually make some decisions on.
Shawn
Got it. What is semi-recurring?
Nikki
Those are, we look at things in terms of the dollar amount and the timing. So something like rent, I pay my rent every month and it's the same dollar amount, that's a solidly recurring. Something like your electric bill is probably gonna vary pretty greatly throughout the year. And so that's a thing where the timing is consistent, but the dollar amount might vary pretty broadly. And that would be semi-recurring.
Shawn
Okay. So one of the things that's always interested me, we've certainly observed that high income people don't necessarily put away the money they should. What I guess they do have is the flexibility that they, you know, when things go worse, they can stop going out, right? They can, whatever, they can cut their Hulu and survive with just Netflix or whatever it may be. They just have more flexibility when the time comes to trim back.
Is that something that you guys empirically see in the data that people actually do pull back? Or are they just kind of always skirting along the edge?
Nikki
Well, I think that that goes a little bit on the assumption that where that money is going is discretionary. And, you know, if we think about the homes people have, yeah, I'm like, we think about the homes people have chosen to be in, sometimes the cars people have chosen to drive, right?
Shawn
Okay, Santa Monica.
Nikki
The the amount of that that is discretionary might not be as much as other people would think. And I think that also it's always an interesting thing. I mean, we've we've all worked in subprime at different times, but that understanding of how do you manage when things get tough? And folks who have subprime credit scores often feel a little bit more resilient because they understand what it's like to stretch a dollar and to figure out how you operate in those ways, where if you're accustomed to high income and then suddenly there's a material change in that situation, you've recently been laid off as the obvious example, you may not have those same structures and you may not have been in that position where you really had to think about it. So how you cut back on those things, how you react to that can look different within a high income population, certainly.
Shawn
That's interesting. You know, the data I want you to find out for me, I say the three leading causes of bankruptcy and default are, you know, loss of income. You lose your job, medical expense, right? Something, something happens there. I think the number three cause is Disney World. So many people that we've talked to in subprime space over the years, you know, have gone to Disney world. And I remember looking at it to go just for a weekend with my daughter was like $5,000. Like it was just a crazy amount of money. So you want to practice your Python, go check that one out for me.
You mentioned consumer permissioned. So when I hear consumer permissioned, I hear friction. So how does this fit into the underwriting flow? Typically with, with you know, a Plaid or any one of those providers, I have to go and provide credentials and some people are anxious about that. Can you talk about that part of the experience?
Nikki
Yeah, absolutely. I think that there are a couple of different layers to that. So the thing that I say as I'm talking to different lenders and clients is no one's ever told me they don't want cashflow underwriting data, right? If you say yes or no, would you like to know what a consumer's depository information is? Everyone's always going to take that. But it is a trade-off because the consumer permissioning does add a level of friction. Absolutely. So to your point, logistically speaking, it does mean the consumer is giving their log in and their password and they've got to go figure out what the password is. Depending on how much they're transacting, we're pulling up to 24 months of data, that takes time. That all looks different than a traditional Bureau report where you're accustomed to in sub one second getting everything it's already nicely packaged. Because as we mentioned before, it's all sitting in an established database and it's already right there ready to go. And so that is a trade off that a lot of our lenders have to kind of think about. And when you think about how you're implementing a credit policy and where you're starting to use the data, you often are going to be a little bit more strategic about where you introduce that friction. So if you are a super prime credit card consumer where people are expecting an instantaneous decision and a very, very high limit, you're probably not going to give someone who has a credit score of 800 the request to log into their bank because they're going to walk away and go somewhere else. So, you do often figure out those places, a very logical place for a lot of people to test is in the marginal approval and decline space. These are folks that I'm just barely approving or I'm just barely declining. And if I start to look at cashflow data, can I cherry pick those? So if it's somebody I'm going to decline anyway, I'm really not losing anything if I say, hey, let me give you a chance to opt in, provide this additional information for me, and I can revisit that decision. You also have to think about how you're speaking to the consumer. So there's very much, a generational gap in terms of my father who is a lifelong accountant and very carefully manages all of his finances doesn't even have online logins for many of his accounts because lives in a small town in Georgia and he's friends with the president of the bank. So he just goes in and talks to them when he wants to. A file cabinet full of paper that I'm gonna have to deal with one of these days. So in those ways, right? Like my dad's a really poor candidate for consumer permission data because he doesn't even have those accounts.
On the other hand, my nephew who's 22 has only ever interacted digitally, right? So that idea of he of course has immediate access to his login and his password and his ability to do that looks very, very different. Additionally, the pandemic changed everything. So what was considered to be friction and a really, really high bar suddenly through the pandemic, almost everybody switched to digital access. Everybody started managing their finances digitally and things like that. And so that ability to understand what that looks like to have access to give permissions. Suddenly there was kind of just this wave of change that hit the industry. And so there are a lot of banks and lenders to talk to that say, Hey, I tried this in 2018 and I could only get 3% of people to choose to do it. That looks very, very different in 2024 because of the pandemic, you know, kind of did in terms of people managing things online as well. So, I think there's a lot of like press that's happening as well. Obviously the CFPB has spoken out about the ability of consumers to own their data and to speak to permission what they do and don't want to share. And I think that there is that push amongst consumers who are thinking about talking about it. There's just a lot more transparency in this. So odds are most consumers much more closely understand what's happening in their checking account than they do what's happening in their credit report. I mean, even as folks who've worked in credit for most of our careers, we probably aren't as good as we should be about checking our credit report every year and understanding what's being reported there. About a third of people have some sort of an error on their credit report. And so understanding what's happening in my checking account actually brings that power back to the consumer to say, yes, I'm choosing to share this. And it's data that I'm intimately familiar with and I'm comfortable being provided to that lender. So it really does shift the onus of what that looks like and the power in those cases back to the consumer to be able to say this is my data and this is how I'd like it to be used, which is from a variety of regulations, the direction everything is going.
Shawn
What you described earlier is what we call iterative underwriting. And I think what we still see, I think with most of the underwriters is that they just have a monolithic process. And so there's frequently this debate of, am I going to go get the checking account data and have 10, 20, 30% drop off? Or am I going to give up on that data? And so what you describe is that iterative approach, which is, as soon as I can approve you, I want to approve you, right? And then I might also ask for it just because I can help set line or adjust terms accordingly. You know, is that a, a C change in the organization or is it just really the most progressive, underwriters or maybe just the newest ones who have built a new system and therefore aren't dependent on kind of the monolithic approach? Are we seeing that change?
Nikki
I think we're getting there. It's a very slow burn, I would say. I think cashflow underwriting has been the next big thing for maybe pushing a decade at this point in time. And so I think that it's something that we struggle with the adoption and figuring out who are the right lenders to continue to push. Obviously very publicly, Peddle, lends only on cashflow underwriting data. So they've kind of set a lot of publicity there in the work that they have done. And so I think that like, as those organizations have been successful, as there's a lot more publicity around the issues that face folks who are no hits or thin files and that ability to lend, it does become more and more a regulatory push from the CFPB and other regulating agencies to say, if there is data with which you can successfully underwrite consumers, you should, right? And that's what Project Reach was intended to do, which was intended to force some of the banks to kind of share some of that information.
And so I think that we'd like to say that a lot of lenders are going to do this because it's great data and because it's the right thing to do for the consumer. But I think that additional shift in the regulatory lens is sort of pushing people to find ways to make these things happen and to maybe get a little more comfortable with what this looks like and what that friction looks like in ways they might not have just entirely done on their own because it was the right thing to do.
Shawn
So it seems to me that in addition to the consumer hurdle, there's a kind of a data hurdle here. How does one make their way into using this data? Do I have to sit and collect it for a year and then let it ferment for a year before I can really figure it out?
Nikki
Yeah, I think that's the base case that a lot of lenders assume going in. And so I think that's part of all of the lens that you approach cash flow should be through the lens of build versus buy. And if you're going to choose to build your own solution and go to an aggregate or get that raw data, build your own attributes, build your own score, that is exactly what you will have to do is accumulate that data over time. You start with a largely biased sample. You then build that and try to create some intelligence, maybe get less and less biased over time and you build that. That process takes years, right? So a lender I previously worked for, we established that. We're working in that space to be able to build that because that was what was required at that point in the market. What I would love for folks to do is come chat with Nova Credit and we like to use the data that we've acquired as we've worked with other lenders to be able to add some intelligence into that. And so there will always be an element of test and learn to be able to understand what your particular population looks like, what are the right thresholds, what are the right margins. But if you're working with a provider who has experience in this decisioning space, that allows you to really jumpstart that cold start problem and be able to move much more aggressively. I know the most important attributes that exist in our data set based on the credit band that you're working with and the product that you're looking at and that ability to start to identify that if there are hard cuts, if you're using our production score, those give you that ability to move much more aggressively into that space to create a very functional strategy early on. And then as you develop some of your own early delinquency metrics and your own intelligence, then you can continue to customize that for your population.
Shawn
Got it. Okay, I think I'm there. I'm still struggling a little bit with, do I integrate this with credit bureau data, the way I would maybe other piece of information, or is this kind of a matrixing approach?
Nikki
A matrixing approach is generally the way that I talk about it with our clients and prospects. I very literally often show a heat map if we overlay the credit bureau data and the cashflow data. What's really interesting is that they're quite orthogonal data sets. So if we think about sort of where that distribution is, it's not completely random, but it is like there's still a lot of places where cash and credit are telling very, very different stories.
And that's the sweet spot. Those are the really interesting swap sets that we start to generate in this space and where we're making a materially different decision once you layer in that cash. And so, I mean, that could be something as simple as using a credit score, your own custom score, a Vantage score, a FICO score, and a production score that uses cashflow data, right? And the combination of those two pieces of information. Not dissimilar in the way that many lenders are already using maybe multiple versions of credit bureau data and maybe some fraud data and other pieces of information to bring together a holistic strategy. So certainly there are lenders who are using that at an attribute level, there's some who are using a score. And so how that works for any given lender depends on what the right answer, what the right constraints are within that lender. But that ability to use both of those and then start to figure out what do you want to do when someone has a high credit score but is overdrawn in their checking account today? Those are very different stories and you've got to figure out how you react to that.
And that's the really interesting part of my job is walking with lenders through that to start to say, okay, if suddenly I told you and I again use one of my personal stories in 2021, I resigned from my prior role. I was finishing up that master's degree. If you look at my credit history, nothing changed in that six months, but there was a period of time where I had chosen to resign and didn't have a paycheck. So if you were looking at my cashflow data, you should have thought about me very, very differently than the way you would have based on my credit score stand alone.
And so those are the sorts of people where you have to start to think about that. Right, somebody who doesn't have a thin pile but has effectively managed their checking account for the past five, six years has paid rent every month. How do you think about that person without a credit score? Because they got burned during the recession, decided they weren't gonna use credit again. They paid off their student loans. And so like, what does that start to look like? And how do you start to think through some of those scenarios?
That's the really interesting part of combining those data sets. And there are a lot of different ways to do it. It really depends on where the organization is in their sophistication. We talked through a lot of like model risk management. How much do you want to kind of get into that? And so we find the right answers based on how that org operates.
Shawn
Yeah, it's interesting. Probably six years ago, I worked with a lender that was pulling that cashflow, trying to do cashflow, primarily cashflow underwriting. And one of the things I had them do is, I think they all picked 10 accounts and they went through the checking account and tried to create signals out of that. And it is interesting how much you can kind of develop a persona out of it. It doesn't mean you're right, but you think you've got a persona.
But I remember the one guy in that organization just fell in love with this one woman who kept taking her balance under a dollar, but never below zero. And just the fact that it gave him great comfort actually, right? That she was managing her money sufficiently tightly, that she was able to avoid those overdrafts. And I'll say similarly, when I was at ZestFinance, all of us in the original team worked customer service. And, I had one customer and she wanted $550. We had $100 increments. I'm like, here's what your payment would be. Can you make that every other week? And she's like, no. And so we took it down to 500. And she's like, well, I really need 550. And I would have given her any amount she asked for because if she really, most people we said, can you make that payment? You could have asked them, can you make a thousand dollar payment every day? And they'd be like, sure, no problem. But she was that precise. And fortunately, I knew how to break this system and actually give her a $550 loan.
All right, so let me wrap up with this question. So you've been doing this for over two decades now. You've played on both sides of the data equation. What should we be looking for in the next five years? What's going to change? And how is it going to change the way we view underwriting and we view the tools that we currently have?
Nikki
Yeah, I think it's gonna be a continuation of what we’ve previously seen, which is this explosion of data sources, right? So we think about things like rental data being reported is becoming much, much more common than it was even just two or three years ago. Buy now pay later is the thing that everybody wants to talk about. How is that kind of upsetting what happens in the credit card space? It has a very generational divide in how it's used. It's not currently reported to traditional databases things like that. So when we think about that, I think the thing that's going to be really interesting is starting to see kind of where that data goes, how prevalent it becomes. And because of things I know from the credit bureau side, there are reasons that those things don't fit into those traditional storage systems. And so there's reasons that things like payday loans and buy now pay later, you can't just shove into the traditional credit setups. And so there will become ways of doing that, ways of looking at that data to ensure that those are still part of that holistic picture.
I think that's the thing that will we just keep painting a better and better picture of the consumer. There's more and more data. We're manufacturing it every single day. And so understanding how we bring all of those together and then what is what does regulation start to do? Because obviously, generally, we're incentivized to be able to lend to as many people, but also in a way that is as safe as possible. And so there's a balance to that. There there are people for whom you should not issue a ten thousand dollar loan. It's not helpful for them. It's not helpful for you as a lender and how do you balance that inclusivity and that desire to grant credit where it is appropriate with this continued flow of additional sorts of information, better and better understanding of who consumers are.
So I think that's both helpful to a lot of folks who might not have traditional credit data in those histories, but also it's going to hurt some people who do have really pristine traditional credit histories and have recently lost a job, for example. And so as we continue to get, you know, better and better technology, the way data science is going allows us to just deal with volumes of data that were unfathomable 20 years ago when I was kind of coming out of undergrad. And so I think about like the models that we used to build and how I would set my computer up to run overnight. And then sometimes I would come back at midnight to check and make sure it hadn't failed. And then I'd pick it back up at 8 am the next morning. So the things that we used to do, right, that computing power is now sitting on this laptop.
And so, how do we start to really digest all of this data, collect more and more of it. I think that's the direction the industry is going and then how lenders use that responsibly, effectively. That's really the interesting part of where we'll be.
Shawn
Yeah, that’s interesting. I do remember seeing people leave paper on top of their computer saying, “Do Not Touch,” because it was running something overnight, but you hit on something. I think there's really interesting, which is it's, it's a more complex world in many senses, right? BNPL is more complex, you know, a lot of the lenders we work with, we advise to withdraw payments on pay days as opposed to, you know, the fifth of the month or whatever the monthly pattern is that the Bureau is kind of set up for. And I understand why that creates complexity for the Bureau. And I understand why people who figure out how to deal with that are gonna do really well. Well, thank you so much for joining us today. I really appreciate it, Nikki.
Nikki
Absolutely, this was a lot of fun. Thank you, Shawn.
Shawn
All right, well, you can follow Ensemblex, you can follow Nikki Cross on LinkedIn, that's N -I -K -K -I. You can also visit us at Ensemblex.com and you can find The Ensemblex Exchange Podcast on all major platforms. Thank you for joining me today and thank all of you for listening.